Enhancing the Effectiveness of Ligand‐Based Virtual Screening Using Data Fusion

Abstract

Data fusion is being increasingly used to combine the outputs of different types of sensor. This paper reviews the application of the approach to ligand-based virtual screening, where the sensors to be combined are functions that score molecules in a database on their likelihood of exhibiting some required biological activity. Much of the literature to date involves the combination of multiple similarity searches, although there is also increasing interest in the combination of multiple machine learning techniques. Both approaches are reviewed here, focusing on the extent to which fusion can improve the effectiveness of searching when compared with a single screening mechanism, and on the reasons that have been suggested for the observed performance enhancement.

In silico prediction of unbound brain-to-plasma concentration ratio using machine learning algorithms

“…The PLS K p,uu,brain direct model (PLSd K p,uu,brain model) has significantly less predictive power (R 2 = 0.41), however, the performance is comparable to the indirect SVM model. Consensus models are often used to obtain improved prediction performance [47]. In the current study, the average value of the predictions from the individual components is used as consensus prediction.…”

Scenarios and Case Studies: Examples for Ligand‐Based Virtual Screening

“…In general, some approaches appear more consistent than others, but as Sheridan and Kearsley said in 2002, trying to identify the optimal method for searching databases is a futile exercise [37]. Instead, research is focusing on what may be the preferred combination to use or how the ranked hit lists should be combined, whether by data fusion [38], parallel selection of individual rankings [39], machine learning and weighting applied to data fusion [40], or application of Belief Theory to data fusion [41]. As this new avenue is being explored, we should be aware that in practice the application of LBVS methods strongly depends on the aim of a given project at a certain time and overly complex and parameterized fusion approaches may not be needed.…”

Computer‐aided (in silico) approaches in the mode‐of‐action analysis and safety assessment of Ostarine and 4‐methylamphetamine

“…This matter will be further investigated in the near future as part of establishing the applicability domain of the in silico target prediction when applied to psychoactive compounds (see Limitations of the study). In some cases, results from two or more different in silico methods would be combined to produce a better result, a method also known as data fusion (Willett, ). However, data fusion only produced a synergistic result in certain circumstances such as the case when the methods are different; that is, one method uses 2D descriptors, and another uses a 3D descriptor (Willett, ).…”

“…In some cases, results from two or more different in silico methods would be combined to produce a better result, a method also known as data fusion (Willett, ). However, data fusion only produced a synergistic result in certain circumstances such as the case when the methods are different; that is, one method uses 2D descriptors, and another uses a 3D descriptor (Willett, ). Given that both algorithms employed the same 2D descriptor, it is better to take the result of one of the algorithm than combining the results.…”